In this video, I'll be telling you about Qwen 3.6 27B, why it looks so interesting for coding agents, how to install it with vLLM, and how you can use it with Hermes Agent, Kilo CLI, and KiloClaw.

--
Key Takeaways:

🚀 Qwen 3.6 27B is positioned as a strong open model for agentic coding and longer-context reasoning.  
⚙️ The easiest serious setup right now is serving the model with vLLM and connecting tools through an OpenAI-compatible endpoint.  
📦 Ollama currently has Qwen 3.6 available, but only the 35B A3B version, not the 27B variant yet.  
🛠️ vLLM makes it easy to wire Qwen 3.6 27B into coding tools that support OpenAI-style APIs.  
🤖 Hermes Agent is one of the most compelling ways to use Qwen 3.6 27B for real agent workflows.  
🔗 Kilo CLI can connect to your local vLLM endpoint through its OpenAI-compatible provider setup.  
☁️ KiloClaw is a natural fit for persistent hosted agent workflows if Qwen 3.6 27B becomes available there.  
🍎 MLX support on Mac, especially Apple Silicon, is worth watching as the ecosystem matures.  
✅ Overall, Qwen 3.6 27B looks like a very promising option for coding agents, especially if you want control, flexibility, and strong tool use.

AICodeKing

Qwen 3.6 27B is a new open model focused on agentic coding workflows, repository-level reasoning, and long-context task tracking. The video covers how to serve it locally using vLLM with an OpenAI-compatible endpoint, and how to integrate it with Kilo CLI, Kilo Claw, and Hermes Agent. Key setup steps include enabling tool-calling flags in vLLM, configuring Hermes via config.yaml with a custom endpoint, and maximizing context window size to take full advantage of the model's strengths. The 27B variant is not yet on Ollama, making vLLM the recommended route for now, with MLX support for Apple Silicon expected soon.

Qwen 3.6 27B + Hermes,OpenCode,OpenClaw: THIS IS SO GOOD! The BEST LOCAL AI CODER!